Measuring the accuracy of facial approximations: a comparative study of resemblance rating and face array methods

The success of facial approximation is thought to depend, at least in part, upon the "accuracy" of the constructed face. However, methods of accuracy assessment are varied and this range in methods may be responsible for the disparate results reported in the literature. The aim of this study was to determine if the accuracy results of one facial approximation were comparable across two different assessment methods (resemblance ratings and simultaneous face array tests using unfamiliar assessors) and if resemblance ratings co-varied with recognition responses. True-positive recognition performance from the facial approximation was poor (21%) while resemblance scores using the same facial approximation were moderately high (3 out of 5 on a five-point scale). These results are not, therefore, consistent and indicate that either different variables are being evaluated by the methods, or the same variable is being examined but with different weight/calibration. Further resemblance ratings tests of the facial approximation to three foil faces from the face array revealed that resemblance scores were similar irrespective of which face was compared, and did not closely correspond with the degree of recognition performance. This was especially the case for isolated comparisons of single faces to the facial approximation. Collectively, these results indicate that resemblance ratings are: (i) insensitive measures of a facial approximation's accuracy; and (ii) inconsistent with results of unfamiliar simultaneous face-array recognition results. These data suggest that familiar and unfamiliar recognition tests should be given increased weight in contrast to current resemblance rating tests.